Decision Tree Analysis in Project Management

My last decision was one minute before I wrote this article. It's still very fresh. This decision was one of hundreds of decisions I made during the day. I made a decision analysis for all of them myself. Of course, these decisions were not big enough to affect the value delivery of a project. If the decisions in question are going to affect the success of a project, it's wise to take a more systematic and scientific approach. decision tree analysis Project management is a technique that involves making decisions that directly affect the success of a project. Decision tree analysis, project managers often use this method, guided by different sources and tools, to make informed decisions.

So, What is Decision Tree Analysis?

Every decision you make has a potential consequence. Decision tree analysis helps project managers analyze these results by providing a representation. In this context. as a project manager you should create a diagram that shows all the possible options, the possibilities and the consequences of each option.  You see the potential risks of every decision you have to make as a project manager this way. You will have an idea of what you can gain and you can act more consciously thanks to it.

When to Use?

As we just said, it is very useful for bringing out the potential risks that may emerge as a result of a decision. It tells you what you can gain as well as the risks. You have made a decision, but what consequences might this decision have on the project outputs? It gives you an approach to find out the answer to this question.And perhaps the most functional feature gives you a choice.  It is useful in the planning performance domain.

How to Apply?

Decision trees are simply a tool to help a business make a decision, and the key inputs into that decision will be the cost of the decision, the probability of success, and the end payoffs—what the financial or monetary outcomes are. Now, the best way to explain decision trees is to use an example, and the example is this: There is a business looking to launch a product, and the launch is going to be one of two ways. It can only be one of these two ways. The first way is they launch the new product using advertising, and the second way is that they launch the new product using sales promotions.

With the start of a decision tree, it will always begin with a square, a box, and that box simply represents the decision—should they launch through advertising or sales promotion? Which way should they launch? You'll then see lines that come off, and they will eventually end at a circle, and these circles represent the courses of action. The courses of action are, of course, advertising or sales promotion.

Now, within each course of action, you will see these variations. And the variations are very crude in this example. You'll either get high demand or low demand. So, if you get high demand from using advertising, because we're on this strand now, so high demand, it is saying 0.7, which represents the probability of that happening if you go with this course of action, which is advertising. So, it's a 70 percent chance, but you write it as 0.7 as we get to later on. And we see here it says the other variation, so the opposite of high demand, that is low demand, and low demand must be clearly 0.3.

And number three plus number seven adds to one, being the full probability range. The last thing you'll see is the final payoffs. So, the final payoffs are what you estimate the attached outcomes if you go with this course of action. If it's advertising and then you are fortunate enough to have high demand, the final payoff is estimated to be 50 million pounds. But if you go with the course of action of advertising and you get low demand, which has a 30 percent chance of happening, then you'll get the final payoff of 20 million pounds.

Just to run through the other side, if we went with the other course of action, sales promotion, then 5 million pounds is the cost. High demand is very likely, 0.9, 90 percent. And the payoffs in that case would be 40 million pounds. And in this case, if it was to be low demand, 10 percent chance, 0.1, and you'd get a 5 million pound payoff if that occurred.

Now, this is very much laid out for you. It might be the case that you have a case study or a context. In that case, you may need to extract the numbers from the case study and create a decision tree. But once you've done that, then you need to apply the formulas. So, the key formulas that you need to be aware of, number one is EV, which stands for the expected value. So, the expected value of one variation, what I mean by that is let's just say the course of action is advertising and the variation that you do advertising and then you have high demand. That would be one variation. So, if you calculate the expected value of that variation, then it would be the final payoff. So, in that scenario, the final payoff is 50 million pounds. So, 50 million pounds times by the probability of that happening, and the probability of that happening, being advertising high demand, is 0.7, 70 percent. So, you would do simply 0.7, the probability times by the final payoff, which is 50 million. So, that comes to 35 million. The 35 million would be the expected value of that individual variation, which leads us into the next formula.

The next formula is you need to calculate the expected value of a particular course of action. So, if you wanted to do that, you would simply take all the variations within that course of action. Now, clearly in this course of action, there are two variations. There is a variation of high demand and a variation of low demand. So, you would simply do calculate the expected value of each variation. So, 0.7 times by 50, and then 0.3 times by 20. And then you would just add them together. If there were three variations, you'd have to add all three together. Before you'd add all four together, so you should sum all the variations. And the last formula you need to know is the net gain. And the net gain is simply, when you've calculated and you've added up all those expected values of those variations, you get to the expected value of the course of action, and you just take away the cost of doing it.

So, in this case, the cost was 10 million pounds. Now, we're going to calculate what the net gain, because that's the key number, what it's going to be for advertising and what it's going to be for sales promotion, to help the business make a decision in this scenario, which way they should go. So, in advertising, if we just do over here before. So, if we calculate the expected value of the variations across here. So, advertising high demand is 50 million times by 0.7, which is 35 million. 0.3 times by 20, looking for a time for 20 is 6 million. We then add them together, so 35 plus 6, 35 plus 6 is 41 million. So, we've done the expected value of that entire course of action. And then we've come to 41 million, and we need to subtract off the cost of doing it. The cost of doing it was 10 million, so that gives us the final net gain of 31 million. So, then getting here is 31 million.

We do the same to the other side. So, for sales promotion, sales promotion, 0.9 times by 40 comes to 36 million. We then do add that to 0.1 times by 5 million, which would be half a million. 36 plus 0.5 gives us the entire expected value of that course of action. So, that is 36.5 million. But we then need to subtract off the cost of that course of action, which is cheaper in this case. It is 5 million, and that leaves us with 36.5 minus 5, which equals 31.5 billion. So, it is a slightly higher net gain on the sales promotion side. So, if the business exclusively decided to use just decision trees to make that decision, it would choose sales promotion because it is marginally higher in terms of its net gain.

Now, the next thing we need to look at is the pros and cons of using decision trees because clearly in this example, it is a very, very fine line. So, if you're thinking about the pros and cons of decision trees, let's start with the pros. So, the first pro is that simply using decision trees is a logical approach. It's a logical approach, using quantitative data that is based on numbers. Quantitative data is objective because you're just going with the decision tree, finding out which has the higher net gain, and that is the correct answer as far as you go. It's clearly more on the scientific side, which may lead to the same advantages you saw there because it could be more justifiable to particular stakeholders, such as shareholders, when you make these decisions.

The last thing to think about is that it does include the costs of each of those courses of actions, so it looks at the risks there. On the con side, number one, it doesn't account for any qualitative data. It only looks at those numbers and nothing else. So if there's a case study question, it's quite likely that there will be some sort of quantitative factors that are in the case value, and you could hook them in as part of your analysis because the decision trees are just going to be only on the quantitative side. So clearly, you want to look at the impact on particular stakeholders. That could be employees; maybe the employees hate the decision, and they're going to leave, and that's going to lead to extra costs like external recruitment costs. Impact on shareholders, maybe shareholders hate the idea and they're not going to invest. Or on customers, and they're not going to buy. Also, you want to look at trade-offs because it might be particularly with one course of action, and that is really myopic. It's a really short-term course of action and doesn't look at the long-term sustainability of the business. Number two, it's only as good as your predictions. If your probabilities are wrong, the whole thing's wrong. If your final payoffs are estimated to be wrong, the whole thing is wrong. So it's clearly based on how good, how accurate they are.

Now, in terms of whether we should use it or not, these are the basics to think about. So yes, you should use it if you've been faced with this decision many times before. If you've done lots of decision trees or you're an experienced business and experienced manager, you've seen the decision over and over again, you're always releasing products. It's always between sales promotion or advertising, then you've probably got a good understanding of it. So in that case, you might just want to go simply with the net gain because you think you know it all. So it's more likely to do a short-term with tactical decisions.

Versus when you shouldn't use it, if it's a brand new type of decision. The reason why, because if it's a brand new type of decision, you probably are going to be likely off your probabilities at final payoffs. How can they be accurate if you haven't seen them before? Number two, if it's a strategic decision. If it's a long-term decision for that business, then you probably want to avoid just solely using decision trees because, as I said before, you want to think about those stakeholders. You want to think about that qualitative data. And number three, the last one is if you have any lack of trust in that data, then you definitely shouldn't exclusively be using your decision trees. So I hope that helps. I'll see you in the next session. 

See also: